The idea of introducing hyperautomation into your company may have you feeling overwhelmed. But it’s easier than you think to pair AI with RPA to get a desirable combination — just like a chocolate peanut butter cup.
In recent years robotic process automation (RPA) has become mainstream, enabling organizations to free their staff from rote, manual processing. At the same time, machine learning (ML) and artificial intelligence (AI) have become more accessible with the advent of off-the-shelf tools, low-code environments, and simplified frameworks.
Each of these technologies is independently reshaping the way businesses operate and compete. Blending them together can have even greater impacts. The resulting capability comprises some of what’s referred to as hyperautomation. That sounds amazing, but it’s not always clear what hyperautomation means or how to go about combining RPA and AI into high-impact, fully automated solutions. In some cases, organizations have launched an RPA program but struggle to incorporate more advanced functionality.
So, how do you get from RPA to AI and beyond?
The Peanut Butter Cup Analogy for Hyperautomation
RPA and AI have a lot in common, like:
- Mimicking human activities
- Automating manual, repetitive tasks
- Reducing the number of human touchpoints in a process
- Increasing processing speed and accuracy.
That said, they also work in distinct ways to solve different challenges. RPA excels at simple, structured and objective tasks. AI is meant for subjectivity, decision making and unstructured actions. In fact, the beauty of these two tools in concert is how well they complement each other towards a common goal.
But what does this have to do with peanut butter cups? Consider the key ingredients: chocolate and peanut butter. Both are delicious on their own, but for different reasons. The combination is tasty, immensely popular and preferred by most to either of the individual components. I think you see where this is going.
RPA plus AI is the peanut butter cup of process automation. Instead of addressing parts of a process with a particular tool, you combine tools to automate more of the process and get closer to no-touch, end-to-end automation. Not only do more steps get automated, but RPA and AI help enable each other to integrate more seamlessly. In other words, the result is greater than the sum of its parts.
A Practical Example for Hyperautomation
To help demonstrate, let’s consider a real-world process and show how it progresses with automation.
In our finance department, we receive invoices from vendors that we need to capture, enter into our Enterprise Resource Planning (ERP) system and process. The invoices are always PDF files but don’t have a consistent format or data and may be scanned or handwritten. With so much variation in the data and presentation, a traditional integration isn’t possible. We have to manually view each file then hand-key them into our system, which is mundane, time-consuming and inefficient. We can do better!
RPA: Start with the Basics
RPA is often the easiest and most accessible entry point for automation. Not only are the tools low-code, but they can provide quick wins at low cost for straightforward use cases. Plus, RPA can be the means to activate AI by, for example, passing data directly to AI services that make decisions and receive structured decisions or outputs to automate further tactical steps. Consider it an ideal starting point and foundation for building towards hyperautomation.
In our example process, we use RPA to automatically receive vendor emails, download attachments and associate those files to the right vendor accounts in our system using the sender’s email address. The RPA solution even creates work items and generates notifications to users so they can promptly respond.
Getting started with AI and ML is probably easier than you think. In fact, chances are you’re already taking advantage of them. Optical character recognition (OCR), where an AI model can read and digitize scanned or handwritten text, is but one ubiquitous example.
There’s even a growing library of pre-packaged “skills” available to add a la carte to your automation. Examples include reading receipts, translating text from one language to another and getting the sentiment of a message. Besides the obvious benefit of quick implementation, these out-of-the-box services are often proven models that work with a higher degree of accuracy.
Incorporating this type of AI, our bot gets two new features – an intelligent OCR model that digitizes PDF text and an invoice extraction model that reads standard data from known invoice elements (like date, amount, and remittance contacts). The RPA automation runs these features on-demand and uses the output to do more precise data entry. As a result, the data now populated in our ERP isn’t simply a document to review but structured data auto-populated in certain fields as well.
Tailored Models to Fit Your Business
As you fold AI into your automated processes, you’re likely to find opportunities where off-the-shelf solutions don’t suffice. For example, our invoice process might need to extract a vendor ID or PO number from each PDF document in addition to the standard invoice fields.
When a basic AI service won’t work, the solution may be “tunable” AI. These services still use pre-built models, so you don’t have to start from scratch, but they allow you to train those models on your specific data and business rules. Training is often simple enough for non-technical users, and the result is AI that’s custom fit but still relatively easy to create.
Using this capability, we train a tunable Intelligent Document Processing (IDP) model to extract the vendor ID and PO number from each invoice on top of the standard data already being captured. Now, our automated process is capturing all the data our system needs and entering it directly into the ERP. Our Finance team just has to validate the entered data and approve it for processing.
To Infinity and Beyond
From here, there are more capabilities to incorporate as part of a fully operational hyperautomation practice. For example, we could add:
- A customized ML model to assess the risk of an invoice and decide to pay it automatically
- Process mining to track process execution and identify bottlenecks with hard data
- Advanced reporting to show the value of automation and reinforce improvements.
All of these contribute to an automation capability that doesn’t only incorporate AI but injects automation and intelligence into every phase of the automation program. Of course, the barrier to entry can increase when adding more advanced features, making the road to hyperautomation – which so far has been relatively flat – get steeper.
Ultimately, though, the journey is more important than the destination. Hyperautomation need not be an all-or-nothing proposition, given that you can achieve operational gains every step of the way. Instead, understand which tools to incorporate in what order and build your capabilities in a more natural way. As our example showed, RPA automation that uses off-the-shelf AI models is still really powerful. And if nothing else, hopefully, you now see RPA, AI and hyperautomation not as imposing technological undertakings but as a peanut butter cup waiting to be eaten.